← Back to team overview

dolfin team mailing list archive

Re: information about multigrid

 

Sorry for the late reply. I have just moved to Chicago and have not been
able to read email until now.

> Dear Anders,
> I would like to ask some questions about Dolfin!
>
> 1:) The multigrid solver you have implemented,
>     can it be used for solving any kind of PDE or
>     just poisson-like problems?

You can try use it with any PDE, but it will probably not converge if the
problem is not Poisson-like...

> 2:) Most of the books about multigrid are
>     finite differences/volumes oriented. Is there
>     any good one for finite elements from which we
>     can undrstand how multigrid is built step by step?

I don't know of a good reference, but I think Erik Svensson
(eriksv@xxxxxxxxxxxxxxxx, I think he's reading this mailing list) knows of
some good references.

> 3:) How much is multigrid performance influenced
>     when performing a renumbering of the mesh?
>     How about Krylov space solvers? (GMRES/CGS/BiCGStab)
>
> 4:) Which preconditioner is better to use for Navier-Stokes
>     combined with which one of the (GMRES/CGS/BiCGStab) for
>     low Reynolds numbers and which for high Reynolds Numbers?

Don't know. The multigrid solver is quite new and experimental. It can
probably be improved in many ways. Tell us if you have any suggestions or
(preferrably) patches. :-)

/Anders





References